Emerging Technologies · By Brian Fulcher
Prototype or demonstration. Xanadu and HyperLight reported lower light loss in an optical switch tested for photonic quantum computing.
Last reviewed November 11, 2026.
A switch that lets almost all the light through sounds good enough. Now put a hundred switches on a route. Each one can take a little more light out of the signal. In a quantum computer that uses particles of light to carry information, a missing photon can mean a missing piece of the calculation.
What the companies measured
On November 10, 2026, Xanadu and HyperLight reported an insertion loss of 7 millidecibels, or 0.007 decibels, for an optical switch made using thin-film lithium niobate. Insertion loss measures how much optical power is lost by putting the component in the path. The companies had reported about 20 millidecibels on July 29, 2025. Both values come from the companies.
Those numbers correspond to roughly 0.16% and 0.46% power loss per switch, respectively. To see why the difference matters, imagine light passing through 100 identical switches with no other losses. About 85% of the incoming power would make it through at the newer figure, compared with about 63% at the older one. That calculation illustrates cumulative loss. A real quantum machine has its own routes, connectors, detectors and sources, each with its own losses.
How the switch steers light
On a photonic chip, narrow tracks called waveguides carry light. An optical switch can change which track the light exits through. Lithium niobate is useful because an electrical signal can change how light travels through it. Xanadu’s 2025 Aurora research paper describes using voltage to change the relationship between light traveling along two paths, so their waves combine more strongly at the chosen exit. The paper provides context for why switch loss matters in the architecture. The November announcement describes a newer component test.
Who could use it
Photonic quantum hardware teams could use lower-loss routing to preserve more of the light their experiments depend on. The companies say their test used a high-volume wafer manufacturing line and report good repeatability. They did not disclose a numeric production yield, the switch’s purchase terms, or results from an integrated quantum computer using the new switches. As of November 11, public access to this specific device and independent verification of the 7 millidecibel result were unconfirmed.
Root Access Take
I’d judge this the way I judge a network link: add up the losses along the route people actually need. A lower-loss switch buys room in that budget. Packaging, connectors and detectors can still consume it. The most useful follow-up would show how much light makes it through an assembled system at the speed its control electronics need.
For a different approach to building reliable quantum hardware, see the IBM and Oak Ridge error-correction explainer.
What to watch next
Look for a technical report with the measurement setup, variation across chips and wafers, and results across a complete route. Switching speed and how cleanly the device sends light to the intended output matter alongside loss. A system experiment would also show whether the change improves the reliability of a protected quantum calculation.
Primary sources
Xanadu and HyperLight’s November 10 announcement · The companies’ July 29, 2025 result · The Aurora team’s 2025 Nature paper.